Twitch Confirms It Was Feeding Streamer Content Into Amazon's AI Training By Default

Twitch confirmed on Wednesday, August 12, 2026, that it has been allowing Amazon to train generative AI models on streamer content, with every account automatically opted in unless the user finds a setting and turns it off.

The toggle, labeled "Training for Generative AI," sits in the Security and Privacy tab of Twitch's account settings, according to Twitch's own account settings documentation. Turn it off and your streams, VODs, clips, highlights, stream chats, and any text or images on your channel are excluded from "future training of a model developed by Amazon whose purpose is to generate or synthesize text, audio, images, or video," per Twitch's FAQ.

But it was already switched on. Streamers didn't get a heads-up or a chance to choose. They got opted in by default and had to stumble onto the setting themselves, according to 404media.co, which reported that this was the case on its own reporter's account and that users across social media reported the same experience.

Content creator and streaming reporter Zach Bussey posted a screenshot of the new toggle on August 12 that went viral, alerting a large chunk of the streaming community to something Twitch had already turned on for them, according to Kotaku. Twitch itself confirmed the rollout the same day via its Twitch Support account on X, framing it as a new option being added for users, not as a disclosure of something already underway.

Twitch's own public messaging said, "We've added a setting that lets you opt out of having your channel content used to train generative AI content models across Amazon." That's true as far as it goes. But it skips the part where the training was already happening, or set to happen, without anyone's consent, as 404media.co pointed out.

Modern platforms run on machine learning for things users actually want, like automated captions, spam filtering, and content moderation at scale. Twitch's FAQ argues some AI features, like AutoMod, can't be turned off individually because "turning those off for one person would weaken safety and security for everyone else on the platform." That's a reasonable operational argument for moderation tools. It is a much weaker argument for generative content models that build new products Amazon can sell.

The opt-out, once you find it, is narrower than most streamers probably assume. Chat text you type in someone else's stream isn't governed by your own preference. It's governed by the channel owner's setting, according to Twitch's FAQ, as reported by both aftermath.site and insider-gaming.com. Opt out all you want. If you chat in a stream where the host hasn't opted out, Amazon can still use what you typed.

The opt-out is also not retroactive. Unite.ai's review of Twitch's documentation notes the exclusion language is built around "future training," meaning anything already folded into a model before a user disables the toggle stays in.

During a livestream addressing the backlash on Wednesday, Twitch's head of community, Mary Kish, told viewers she personally opted out of AI training on her own account, according to Kotaku. She argued Twitch "can't function without some forms of AI," while acknowledging AI tools are now embedded across many streaming communities. Twitch's chief product officer, Mike Minton, appeared alongside her and cited a study claiming roughly 50% of adults now use AI tools like ChatGPT, framing broad AI adoption as a reason to keep training running. Minton previously said in 2024 that Twitch content was already being used to prototype Amazon's generative AI models, though not yet "in any kind of production scale," according to 404media.co.

IGN's Bo Moore noted that the top request on Twitch's own User Voice feature forum, posted back in May, asks the company to "make all AI features optional, and opt-out by default." That single suggestion has pulled more than 13,000 votes and 4,000 comments, according to IGN, dwarfing every other request on the board by a wide margin.

None of the coverage here suggests Twitch did anything illegal. Nothing in these reports points to a lawsuit, an FTC complaint, or a regulatory inquiry tied to this rollout. This is a product design and disclosure fight, not a legal one, at least so far.

The open question is whether streamer backlash forces Twitch to flip the default, the way public pressure has made other platforms reverse opt-out AI settings. For now, if you stream on Twitch and don't want Amazon training AI models on your face, voice, and chat logs, you have to go find the switch yourself.